【作者】 梅玉洁；

【导师】 石永革；

【作者基本信息】 南昌大学 ， 计算机应用技术， 2006， 硕士

【摘要】 由于网络规模的不断扩大，网络管理研究成为IT业界的热门研究方向之一，通信运营商急需建设综合化、集中化、智能化的网管系统，对通信网进行有效的管理，为客户提供优质的服务，提高企业的竞争力。 通信网综合网管需要采集和处理来自交换、传输、数据、动力环境等专业网的海量告警数据，为了保证有效地防止告警风暴，保证关注告警，快速发现、定位和处理故障，必须设计和实施有效的告警过滤机制，以实现快速、准确地过滤瞬断告警、无效告警和次要告警，这是建设通信网综合网管的核心任务之一。 本文首先对通信网告警数据的特点进行了详细分析，并详细阐述了告警关联分析的常用方法以及关联规则挖掘算法。 其次，在比较现有告警关联分析方法后，探讨了如何采用数据挖掘技术获取告警关联规则，并在此基础上构建告警过滤规则知识库，以及各告警过滤模板的知识表示和知识获取更新方法。 最后，以上述研究分析结果为基础，提出了分层告警过滤模型和基于模板的告警过滤方案，阐述了知识库规则匹配及冲突消解方法，告警过滤机制的具体实现及其实际应用效果。 国内对于本地通信网综合网管的研究与开发工作是近几年才起步的，而将数据挖掘和知识库技术应用于告警过滤机制的研究工作则时间更短，因此本课题的研究内容与成果具有较强的理论、技术与应用价值。课题的主要创新体现在： ◇ 根据对告警数据特点和告警过滤需求的分析，定义了分层结构的告警过滤模型。 ◇ 设计构建了告警关联知识库，实现了告警过滤模板的智能化生成。 ◇ 通过分析经典Apriori算法并结合时间窗口，获得告警时间相关性规则并应用于实际系统。 ◇ 采用了基于模板的复合过滤技术，有效地实现了告警数据的逐层收敛、告警过滤功能的灵活配置与实施。更多还原

【Abstract】 With the continual expansion of the scale of network, network management research has become one of the most popular research directions in IT industry. Telecommunication service providers need to construct a integrated, centralized, intelligent management system, in order to carry through effective management to telecommunication, to provide qualified services to customers and to improve the competitiveness of enterprises.Telecommunication integrated network management demands to collect and process the huge number of alarm data from the exchange network management, transmission network management ,data network management, dynamic environment network management and etc. It’s one of the core tasks that designing and implementation effective filtering mechanisms of alarms to achieve filtering disrupt, void and secondary alarms rapidly and accurately, to ensure to prevent alarms storm effectively, to ensure the concern alarms and to detect, position and handle failure rapidly.Firstly, this article analyses the characteristic of the alarm data of telecommunication network in detail and minutely clarifies the common methods of alarm relativity analysis and association rule mining algorithm.Secondly, it probes into how to obtain the alarm association rule using data mining technology after the comparison of current alarm association analysis methods. On based of that this article constructs alarm filtering rule knowledge base and knowledge expression of every alarm filtering module and knowledge acquisition and updating methods.Lastly, based on the former analysis results, this article puts forward stratified alarm filtering model and alarm filtering project basis of module. Further more, it illuminates knowledge base rule matching and the methods to settle the collisions, the concrete implementation of alarm filtering mechanism and its application effects.In our country, it is not a long time to start the research and exploitation of local更多还原